{"id":"lfm3-1-7b","slug":"lfm3-1-7b","name":"LFM3-1.7B","orgId":"liquid","family":"lfm3-1-7b","releaseDate":null,"availability":"announced","openWeights":true,"license":null,"paramsTotal":1700000000,"paramsSource":"slug","contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://www.liquid.ai/blog/lfm3","hfRepo":null,"modalities":["text"],"aliases":["LFM3-1.7B","lfm3 1.7b","lfm3-1.7b","liquidai/lfm3-1.7b","lfm3-1.7b-base","liquidai/lfm3-1.7b-base","lfm3 1.7b base"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-27","announced":"2026-10-27","blurb":"Liquid AI's LFM3 text model for local agents and specialized applications, on a new tri-hybrid architecture of gated short convolutions, sliding-window attention and global attention, with a 256K-token context. Open weights under the LFM Open License v1.0, with the pre-trained LFM3-1.7B-Base released alongside.","api":"live","openWeights":"yes","params":{"text":"1.73B","source":"Hugging Face model card (safetensors: 1,728,801,536)"},"facts":[{"label":"Architecture","value":"24 layers: 13 gated short convolutions, 5 sliding-window attention (256-token window), 6 global attention"},{"label":"Context","value":"256K tokens (input + output)"},{"label":"Training","value":"22T tokens of pre-training (last 3.5T curated mid-training with agentic traces); SFT, mixed RL, multi-domain on-policy distillation and agentic RL"},{"label":"Languages","value":"English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, Polish"},{"label":"Memory","value":"About 1 GB at 4-bit precision, per Liquid"},{"label":"Checkpoints","value":"Post-trained LFM3-1.7B and pre-trained LFM3-1.7B-Base on Hugging Face (plus GGUF, MLX and ONNX builds)"},{"label":"License","value":"LFM Open License v1.0"},{"label":"API","value":"Liquid AI API and Playground, per Liquid's blog"}],"claims":["Lab claim: outperforms LFM2.5-2.6B with 35% fewer parameters, averaging 49.37 on Liquid's agent, tool-use, coding, instruction-following, knowledge and long-context table vs 47.24 for Qwen3.5-9B and 39.27 for Qwen3.5-4B.","Lab claim: at 32K context (llama.cpp, Q4_0) it prefills at least 2.5x and decodes at least 1.8x faster than MiniCPM5-2B, Qwen3.5-4B, Qwen3.5-9B and Gemma 4 E4B on three devices, using the least memory."],"sources":[{"kind":"lab","title":"Liquid AI: Introducing LFM3","url":"https://www.liquid.ai/blog/lfm3","date":"2026-10-27"},{"kind":"lab","title":"Liquid AI announcement on X","url":"https://x.com/liquidai/status/2115128329003303107","date":"2026-10-27"},{"kind":"repo","title":"LiquidAI/LFM3-1.7B on Hugging Face","url":"https://huggingface.co/LiquidAI/LFM3-1.7B"},{"kind":"repo","title":"LiquidAI/LFM3-1.7B-Base on Hugging Face","url":"https://huggingface.co/LiquidAI/LFM3-1.7B-Base"}],"firstBenchmark":null},"organization":{"id":"liquid","name":"Liquid AI","slug":"liquid","country":null,"url":null,"color":"#1F6FEB","mark":"L"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 27, 2026; no independent results yet.","lead":"As of Nov 3, 2026, LFM3-1.7B is pending benchmarks on isitsota: announced Oct 27, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}